Dissertação de Mestrado

GiTab: Automatic Tablature Music Composition for Classical Guitar

Martim Zanatti dos Santos Gomes da Silva2021

Informações chave

Autores:

Martim Zanatti dos Santos Gomes da Silva (Martim Zanatti dos Santos Gomes da Silva)

Orientadores:

Helena Sofia Andrade Nunes Pereira Pinto (Helena Sofia Andrade Nunes Pereira Pinto); David Manuel Martins de Matos (David Manuel Martins de Matos)

Publicado em

23/09/2021

Resumo

A guitar tablature is a form of musical notation that indicates the fret and string on which each note should be played. For beginner guitar players and guitarists who cannot read scores, this form of notation is essential. On the guitar, unlike other instruments such as the piano, each note exists in one or more different positions on the guitar. So, when we want to play a set of notes, there are several ways to do it. The goal of the thesis is to produce automatic tablatures of classical music by trying to optimize the positions chosen for each note. It is also a goal of this dissertation to choose the fingers that press the frets in order to produce the sounds of the respective notes. The problem of automatically generating a tablature by optimizing the choices of positions is a typical AI problem. Since there are one or more different positions for each note, the growth of possible ways to play a set of notes is exponential. We create the GuiTab system which, using a set of rules and heuristics, can optimize the choice of positions and exclude the least promising hypotheses. The results of our system were compared with published tablatures, where we obtained an average of 68\% accuracy, values like those obtained by works based on Machine Learning. The subjective evaluation showed us that the tablatures generated by the system are always playable, and sometimes the system choices that diverge from published tablatures are justified.

Detalhes da publicação

Autores da comunidade :

Orientadores desta instituição:

Domínio Científico (FOS)

electrical-engineering-electronic-engineering-information-engineering - Engenharia Eletrotécnica, Eletrónica e Informática

Idioma da publicação (código ISO)

eng - Inglês

Acesso à publicação:

Embargo levantado

Data do fim do embargo:

28/07/2022

Nome da instituição

Instituto Superior Técnico